Thermocouple Calibration Method 2

[EmanueleS]

Dear all,

I just would like to ask you for some advice.

I have to do some temperature measurements with a thermocouple K type and then acquire the voltage output with a NI DAQ card.

I have got an analog conditioner which compensates the cold jucntion of the thermocouple, linearizes and amplifies the signal. Of course, the use of this device requires a calibration of the thermocouple system. I thought that this was to best way to treat the output signal from the thermocuple before acquiring it with the DAQ card.

However I was told that this system is not so useful, as I can put the cold junction of the thermocouple in bath ice+water, which is always 0 C and then read directly the output, deriving easily the temperature from the thermocouple emf-T table for rhe K type, which is standard for all the K-type thermocouples.

Do you think this method (which seems quite outdated to me) would be good also if I have to acquire the signal with a DAQ card?

Please let me know.

Regards,

Emanuele

 

[btrueblood]

The signal conditioner for thermocouples is also sometimes known as an artificial ice point, i.e. it replaces the ice/water bath. But, the ice/water bath is a useful double-check that all is working correctly, and can be used to calibrate the thermocouples if the water is pure enough (i.e. dip the probes into an ice bath and verify that the signal conditioner is putting out a 0 deg. C signal). The third method is to measure the temperature of the thermocouple leads at the NI-DAQ inputs, using another device (thermistor, RTD, silicon temp. sensor), then (assuming both legs of the thermocouple and the temp. sensor are all at the same temperature), using the thermocouple emf tables to determine the temperature at the thermocouple junction (business end of the probe).

 

[EmanueleS]

Dear btrueblood,

thank you. So you recommend using the electronic signal conditioner, don't you?
I don't unserstand why people sometimes use the other method, because in my view:

1) the calibration law is not linear
2) the signal is very low (of the order of mV) and so it may be affected by external noise, giving a lower accuracy measurement
3) I don't think it's safe to put wires in a ice-water bath and then plug it into a computer

Don't you think the signla conditioner is a better solution nowadays?

Thanks again

Emanuele

 

[btrueblood]

Well, do realize that the linearization circuit (signal conditioner) is linearizing an inherently non-linear device, so the signal conditioner is a source of error. Typically, the signal conditioners work best over a narrow range of temperatures (relative to the useful range of the thermocouple itself), in order to keep the error in temperature reading low. If you are measuring a very large range of temperatures with a single probe, then you may want to consider other methods. Point #2 is correct, but there are ways to combat noise problems (shielding, filtering). For point #3 - you don't need to put bare wires into the ice water, you could put a sheathed probe into the bath instead; thermocouples are inherently low voltage devices, so the risk of electrocution isn't there (or at least shouldn't be there - I am reminded of an engineer who put thermocouple probes directly onto high voltage wiring and had a very shocking experience - but no ice bath was involved either).

When it comes to temperature measurement, every engineer is going to have his/her opinions. Use what you want, but understand that every choice has its drawbacks. What is most important, I think, is that whatever method you use you should have a good understanding of the sources of error in your measurements, and report those error sources in your report/analysis. And an end-to-end calibration of your thermocouples, both before and after testing, is always a good idea.

 

[IRstuff]

Why not a PRT approach?

TTFN
faq731-376

 

[EmanueleS]

thank you for your message btrueblood.
IRstuf, I'ma afraid but i don't know what a PRT apporach is...

 

[IRstuff]

Platinum resistance thermometer:

http://www.isotechna.com/TTI_22_Standadrd_Thermometer_0_001_C_p/tti22.htm

http://www.hotektech.com/PRST501.htm

Note in the second site, the thermocouples are about about 10x less accurate. The first site claims 1 mK accuracy.

That's assuming, of course, that the sensor is calibrated with the instrument.

TTFN
faq731-376

 

[EmanueleS]

Yes, I think it's fine, but for my application I've already bought three thermocouples and I would like to use them...My temperature range is between 15°C and 40°C, therefore it is a common range.
I just would like to know what calibration procedure is more accurate or advised especially if I have to acquire digitally the signal afterwards...
So I thought that nowadays the method of using the non-linear relation related to the K calibration wasn't used any more, but according to btrueblood it seems that it's still used a lot, of course specifying the error sources, isn't it?

 

[btrueblood]

Emanuele,

In the temperature range you are interested in, it will likely be a wash. In other words, the error due to the linearization of the thermocouple curve (error in the electronic signal conditioner) is likely less than the inherent errors in the thermocouple wire itself.

Calibration is the comparison of a measurement instrument to known good (or better) measurements. Calibration of your thermocouples in that temperature range is pretty simple - get some distilled water and freeze some of it into ice cubes. Make a distilled water + ice bath, and dip your probes into it, and record a few minutes worth of data with your DAQ, then take the same water and boil it and take some more data samples. Knowing the ice is at 0 C (within a few tenths of a degree), and the boiling water is at 100 C (minus a bit due to altitude and possibly barometric pressure variations - see

http://www.engineeringtoolbox.com/boiling-points-water-altitude-d_1344.html),

and the errors in the known temperatures should be a fair bit less than the inherent uncertainty of the thermocouple wire, you have just calibrated your thermocouples.

Regarding PRTs (or platinum RTD's), thermistors, and solid-state or silicon temperature devices, as I said earlier - everybody has their favorite device. What you use depends more upon what limits of error you need, and how much money you have to spend. If you are poor, or have the time, doing your own calibration saves money, but also gives you a much better idea of what the error in your measurements is, and why the error exists. But that too, is just one engineer's opinion.

 

[MatthewL]

btrueblood, my thoughts exactly re using FP/BP pure water for calibration curve. Star for you.

Regards,

Matt

Quality, quantity, cost. Pick two.

 

[HDS]

You can also get a thermocouple module for your NIdaq that does all the compensation and look up table for each kind of thermocouple.